2
\$\begingroup\$

Old analog telephone circuits using carbon microphones seem to be said to have a frequency response of, from various low frequencies which differing reports put at between 275 Hz and 400 Hz, up to various high frequencies, reportedly between 3 and 3.4 kHz, with possibly a peak a around 1 kHz, depending on the manufacturing decade of the carbon mic, and otherwise not that flat.

Does anyone have a reference for some actual audio frequency response curves, including the shape of the peak(s) if any, and the steepness of the high and low frequency roll-offs, for these old analog (POTS) telephone circuits?

\$\endgroup\$
1
\$\begingroup\$

Carbon mic's were pressure controlled resistive devices that produce variable current when driven by a voltage source. They had poor frequency response but the side tone could be cancelled out or adjusted to a minimum easily, so your own voice was minimized on your earpiece. (sidetone) But the freq response was as it is defined 400~3.2Khz today. with 8kbps digital rates present day telephony is limited by a 7th order "brick filter" at 4KHz to satisfy requirement to block all image signals> 4kHz.

I do not have curves for skirt steepness on carbon mics.

\$\endgroup\$
  • \$\begingroup\$ I see recently some pop rock singers use old style portable mics with old-school narrow-band nasal phone sound effects. \$\endgroup\$ – Sunnyskyguy EE75 Jul 18 '12 at 13:34
1
\$\begingroup\$

It's not just the microphone. The frequency response of the entire network was determined quite early on by Bell researchers trying to understand the necessary bandwidth to make conversational speech understandable. Somewhere in this process was i think when they invented the decibel and parent unit the bel, designed to quantify audio levels in a way that was intended to consider perceived loudness rather than source power levels. The system they came up with is approximate but the idea is that a one-dB change should be detectable by a careful person, a three-dB change easily noticeable, and a ten-dB change representing perceived volume of double or half. In terms of actual power, 3 dB difference doubles or halves the power, and 10 dB or one Bel by a factor 10 of ten or 1/10th. They determined that for reliable recognition of a voice and clear understanding it was necessary to have a system b/w of about 3kHz, extending from around 350 to around 3500 Hz. This or a variation was as the system spec for the entire Bell system and is what gives you the typical "telephone voice". What's outside those ranges represents the difference; basically better timbre below 400 Hz and better articulation of consonants above the upper limit.

\$\endgroup\$
0
\$\begingroup\$

As dbeierl says, it's not just the mic (although the mic will be specced no better than it needs to be), it's the entire system. These days (in fact, since the dawn of digital exchanges) that means at the exchange end there is an ADC sampling at a given rate / depth.

For Marconi System.X and Ericsson AXE-10 (the two main systems in the UK network) and presumably almost all others, one analogue phone line maps to one PCM channel in an E1 link, which is 64kbit/s. Obviously there is some analogue loss/distortion in the network & system so you never see that full rate (hence why even 56k modems never hit the full 56k - even when plugged directly into the exchange line card, yes, we tried it) and hence why an ISDN link is 64k/channel.

Mobile phones have even lower bandwidth but use more cunning encoding etc.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.